Abstract

RESULTS

Ts65Dn mice are trisomic for ~16 Mb of mouse Chromosome 16, containing ~105 Hsa21 orthologs, from Gabpa to Znf295 (Akeson et al. 2001; www.ensembl.org). We studied well-supported and annotated genes (known genes or transcripts with multiple ESTs) that could unambiguously be mapped to Hsa21. Mouse orthologs of Hsa21 genes were identified by reciprocal BLAST searching (Reymond et al. 2002b) and by use of mouse genome resources available via the ENSEMBL project (www.ensembl.org). This produced a list of 104 genes and, after excluding the cluster of 22 keratin-associated proteins (Krtap) genes, we tested the remaining 82 genes. An efficient Taqman assay was obtained for 78 of these genes. We also included 15 control genes, six of which are Hsa21 orthologs, but are not triplicated in Ts65Dn, and six normalization genes (Methods).

A total of 99 genes were compared between Ts65Dn (Ts) and euploid (Eu) mice in two developmental stages (postnatal day 30 and 11 mo) and six different tissues (brain, heart, kidney, liver, lung, and muscle). For each tissue under study, equal quantities of RNA from four individual male mice were pooled. Ts65Dn mice, although not isogenic, are a mixture of only two strains (C57B6 and C3H; www.jax.org), so the level of genetic variation is considerably less than that in humans. Pooling of individual mice, while reducing the variation seen, enables aneuploidy-related differences to be identified without being confounded by experimental or inter-individual variation.

Example amplification plots are shown in Figure 1, and give an indication of the precision of the five replicates. Overall, the standard errors (SEM) of Ts/Eu expression ratios are low, 85% genes have a SEM <0.05 (the standard errors for all genes are given in Supplemental Fig. 1). The use of qPCR enabled the quantification of almost all genes in the study; on average 86% were detected for a particular tissue (Table 1). This is considerably better than microarray platforms, where generally 30%–60% of genes are detected, depending on the tissue (data not shown).

Open in a separate window

Figure 1

Example of Taqman real-time PCR amplification plot from P30 kidney of Ts65Dn and euploid mice. Two test genes (N6amt and C21orf18) and one normalization gene (Ppia) are shown. For each mouse genotype and gene there are five replicates. (Inset) The normalized relative expression of Ts65Dn to euploid (xΔe) and the standard error (SEM).

The results for all genes and tissues are given in Figure 2 and summarized in Table 1. Expression results are represented as the normalized relative expression ratio Ts/Eu (xΔe). On the basis of the 95% confidence interval of each value, Figure 2 also indicates how genes are expressed compared with the ratios 1.0 and 1.5. The mean for the control genes is close to 1.0 for all tissues, demonstrating that the normalization strategy we used is reliable. For the triplicated test genes, the mean is close to the theoretical value of 1.5-fold, suggesting that the increase in DNA copy number leads to the expected increase in RNA levels. However, the data are significantly skewed (g₁ = 2.70, P < 0.001; Sokal and Rohlf 1995); many genes are expressed at a level significantly different from 1.5-fold in each tissue (Figs. ​(Figs.2,2, ​,3).3). On average, 45% of the genes are expressed at significantly lower than 1.5-fold, and 9% are not significantly different from 1.0 (Table 1). Strikingly, 18% of the trisomic genes are expressed at >1.5-fold. Thus, only approximately a third of the genes (37%) are expressed at the theoretical value of 1.5-fold.

Open in a separate window

Figure 2

Normalized relative expression Ts/Eu (xΔe). The values are the expression of a gene for Ts65Dn relative to Eu for the tissues and stages shown. The colors represent expression relative to the theoretical values 1.5 and 1.0, based on 95% CI for the values.

Open in a separate window

Figure 3

Histogram of normalized relative expression Ts/Eu (xΔe). The values from Figure 2 are plotted into bins differing by 0.05. The dotted line indicates the position of the value 1.5.

Although the majority of genes are expressed between 1.0- and 1.5-fold, for the molecular pathology of DS, it is interesting to identify genes that have expression ratios significantly different from these levels (Fig. 2). Genes that are consistently expressed at higher levels in many tissues, for example Adamts1 and Mx1, may be considered candidates for aspects of the DS phenotype. One gene in particular, Ankrd3 appears as a candidate for the heart defects seen in 40% of DS individuals (Epstein 2001), as it is highly overexpressed in both P30 (2.83-fold) and 11 mo (5.09-fold) heart and maps within the DS heart critical region (Barlow et al. 2001). Although Ts65Dn mice do not have the heart defect, Ankrd3 overexpression in humans in combination with genomic variation outside of Hsa21 may be causitive. Conversely, the 9% of genes that are not significantly different from 1.0 are less likely to contribute to DS phenotypes.

The coefficients of variation (CV, a measure of the variation independent of the mean; Sokal and Rohlf 1995) of the test and control genes are significantly different (t-test, P = 0.015); this indicates that expression of the triplicated genes is more variable than the control genes. This difference, and the fact that the majority (63%) of genes are not expressed at 1.5-fold, suggests that there are considerable differences in gene responses to aneuploidy. If this variation in expression was stochastic, we would expect that there would be no correlation in gene expression when comparing different tissues. Plotting xΔe for the genes (Fig. 4), indicated that there might be a similar expression pattern between the different tissues and stages. To test this, correlations in gene expression between different tissues were calculated (Table 2) and show that the fold expression in the same tissue at the two different developmental stages are highly correlated. Many of the other tissues also show a high degree of correlation in gene expression (Table 2). In contrast, for the control genes, only five (of the total of 45 pairwise comparisons) showed a significant correlation, and r was very small (mean value 0.01, data not shown), indicating that the variation observed in the control genes was due to either experimental variation or natural variation in gene expression. In summary, the data indicate that expression is not disrupted in a random manner, but is reproducible for most genes across most tissues.

Open in a separate window

Figure 4

Graphs of normalized relative expression Ts/Eu (xΔe). Points represent the values from Figure 2 and bars represent the 95% CI of these values. The x-axis has genes plotted along mouse Chromosome 16 for the test genes. The y-axis represents xΔe. The list and order of genes for each tissue is given in Figure 2.

DISCUSSION

The small differences in gene expression measured in this study are not easily detectable by other methods of RNA quantification. Other studies have used microarrays and serial analysis of gene expression (SAGE) to study global gene expression changes in DS (FitzPatrick et al. 2002; Mao et al. 2003) and Ts65Dn (Chrast et al. 2000; Saran et al. 2003). Whereas these techniques can distinguish between trisomic and euploid individuals by considering global expression-pattern differences, they are poor at assessing the effects of trisomy on the level of expression of individual Hsa21 genes. Real-time quantitative PCR has considerable advantages over these methods in that it can be used to accurately detect very small changes in expression, and the sensitivity means that expression values can be obtained for almost every gene of interest. qPCR thus has advantages over arrays to measure gene expression in aneuploidy and to measure normal expression variation as a quantitative trait (Cheung and Spielman 2002; Cheung et al. 2003; Schadt et al. 2003).

The data presented reveal that aneuploidy results in overexpression of the genes on the aneuploid chromosome. It is difficult to predict how changes in DNA copy-number affect gene expression relative to the diploid state. This has relevance for both studies of aneuploidy and transgenic studies, as the assumption of a direct relationship between DNA copy-number and relative RNA expression may not hold.

Down syndrome results from a complex interaction of genetic, environmental, and stochastic variation (Shapiro 1997; Pritchard and Kola 1999; Epstein 2001; Reeves et al. 2001). The data presented here suggest that expression from the aneuploid chromosome is similar across different tissues and developmental stages. This contrasts with the disruption in global gene expression noted in microarray studies (Chrast et al. 2000; FitzPatrick et al. 2002; Mao et al. 2003; Saran et al. 2003). It thus seems that there is an amplification of the initial variation that may, in turn, result in the extensive variability of the phenotype that characterizes DS. Our data show that aneuploidy results in a subtle and complex overexpression of genes from the aneuploid chromosome.

METHODS

Acknowledgments

Notes

Article and publication are at http://www.genome.org/cgi/doi/10.1101/gr.2090904.

Footnotes

References

WEB SITE REFERENCES